theTinker Closed Account
#1

Back about 10 years ago. I believed i had a decent machine operating at about 2.1ghz. Then i think celeron processors came out pushing this up to really high numbers like 4-5ghz cheaply enough.

However celerons were rubbish from what i was told, but they had a high ghz speed.

Today. I use a 2.6ghz i7 920..

Where have all my ghz gone? It seems like a MASSIVE slow down occurred in growth. Its nice having a quad core etc, but when im compiling some code or using 1 program. I want it done fast as possible.

Can anyone point me to an article that explains this slow down of growth? and why my CPU speed is the same as it was 8 years ago? (obviously im missing something)

AlmightyCushion Moderator
#2

Even that i7 using just one core will be much faster than a Pentium 4 at 3.8Ghz or higher. The reason being that the architecture is a lot more efficient. The i7 processor can do more than the Pentium 4 with a lower clock speed(Ghz in this case).

GarIT Registered User
#3

theTinker said:
Back about 10 years ago. I believed i had a decent machine operating at about 2.1ghz. Then i think celeron processors came out pushing this up to really high numbers like 4-5ghz cheaply enough.

However celerons were rubbish from what i was told, but they had a high ghz speed.

Today. I use a 2.6ghz i7 920..

Where have all my ghz gone? It seems like a MASSIVE slow down occurred in growth. Its nice having a quad core etc, but when im compiling some code or using 1 program. I want it done fast as possible.

Can anyone point me to an article that explains this slow down of growth? and why my CPU speed is the same as it was 8 years ago? (obviously im missing something)


Newer processors are more efficient and can produce more processing power from less Hertz, we could have 100ghz processors but they would be a lot bigger as well as being way more expansive to run. Would you like a laptop that lasts 5 mins on battery, or a desktop that costs €100 per hour to run.

theTinker Closed Account
#4

I take your points. I know that in a single clock cycle the new processors get more work done(well ive read that anyways).

But where is all the speed going?


Loading a weblogic domain takes a full minute or 2 to start up.
Compiling java and copying files/deploying. Takes far too long.

The biggest bottle neck i see with compiling is always the CPU maxed out.
Does Javac use multi cores or is it just using a single core?

When loading a program up after clicking the exec. where is the bottle neck there? (eg: a large IDE editor, or application server etc)

I imagined 2012 being more star treky!

1 person has thanked this post
AlmightyCushion Moderator
#5

Depends on the program and task really. E.g games are usually limited by the graphics card. Hard drives are a big bottleneck at the moment, moving to an SSD makes a big improvement to a system, whether it'll improve it the way you want it to is another story though.

GreenWolfe Registered User
#6

theTinker said:
I take your points. I know that in a single clock cycle the new processors get more work done(well ive read that anyways).

But where is all the speed going?


Loading a weblogic domain takes a full minute or 2 to start up.
Compiling java and copying files/deploying. Takes far too long.

The biggest bottle neck i see with compiling is always the CPU maxed out.
Does Javac use multi cores or is it just using a single core?

When loading a program up after clicking the exec. where is the bottle neck there? (eg: a large IDE editor, or application server etc)

I imagined 2012 being more star treky!


http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6629150

Looks like Javac is single threaded (that particular bug hasn't been closed), you won't get any extra performance from more cores when you're compiling. Post explaining sun bug states here.

You might have answered part of your own question there. Afaik when you compile a Java project, lots of files will be compiled, created and deleted. Disk performance will suffer when doing operations on lots of little files, compared to one big file. You can see this when you copy lots of small files, and compare it to copying a single file of the same size.

maki Registered User
#7

GarIT said:
Newer processors are more efficient and can produce more processing power from less Hertz, we could have 100ghz processors but they would be a lot bigger as well as being way more expansive to run.


We really couldn't. Transistor gate length is a major limiting factor and we're approaching the upper bounds of Moore's Law. To be honest I don't think we'll ever see clock speeds over 10GHz, at least not on silicon.

Torqay Closed Account
#8

theTinker said:
Where have all my ghz gone?


Your ghz are nothing but a sales pitch (back then, as they're now), compare benchmarks before you buy.

Karsini Subscriber
#9

Intel had major trouble getting the Pentium 4 over 4 GHz, the heat and power consumption were just too much. After that, they began working on making processors more efficient at the same speeds and using more of them (multi-core processors).

1 person has thanked this post
GarIT Registered User
#10

maki said:
We really couldn't. Transistor gate length is a major limiting factor and we're approaching the upper bounds of Moore's Law. To be honest I don't think we'll ever see clock speeds over 10GHz, at least not on silicon.


Would it not be possible to make a processor say as big as your computer now, and just add way more transistors?

maki Registered User
#11

GarIT said:
Would it not be possible to make a processor say as big as your computer now, and just add way more transistors?


Simply adding more transistors doesn't make the chip faster. The bigger the chip, the longer the path length, the slower it gets. Not to mention something like that would be impossible to cool. It would literally just melt immediately.

GarIT Registered User
#12

maki said:
Simply adding more transistors doesn't make the chip faster. The bigger the chip, the longer the path length, the slower it gets. Not to mention something like that would be impossible to cool. It would literally just melt immediately.


I was under the impression more transistors would make it more powerful, that's why I do software Theoretically if you kept making it bigger and adding transistors although it would get slower would the hertz not get higher? And supposing you stuck it in the cooling system in a nuclear reactor it would work. Obviously its theoretical and a mad idea.

AlmightyCushion Moderator
#13

GarIT said:
I was under the impression more transistors would make it more powerful, that's why I do software Theoretically if you kept making it bigger and adding transistors although it would get slower would the hertz not get higher? And supposing you stuck it in the cooling system in a nuclear reactor it would work. Obviously its theoretical and a mad idea.


More transistors will make it more powerful but it won't affect the clock speed. The clock speed is the speed of the transistors whether you have 1 or a billion or a trillion. A processor that operates at 3Ghz means it's transistors are operating at 3Ghz no matter how many there are.

GarIT Registered User
#14

AlmightyCushion said:
More transistors will make it more powerful but it won't affect the clock speed. The clock speed is the speed of the transistors whether you have 1 or a billion or a trillion. A processor that operates at 3Ghz means it's transistors are operating at 3Ghz no matter how many there are.


That's where I was getting things wrong, I knew hertz was a measure of frequency but I thought it would be collectively, i.e. two transistors at 1hz = 2hz not 1hz.

GreenWolfe Registered User
#15

GarIT said:
That's where I was getting things wrong, I knew hertz was a measure of frequency but I thought it would be collectively, i.e. two transistors at 1hz = 2hz not 1hz.


No. A 2GHz dual-core CPU is not marketed as a 4GHz CPU....despite what you'll see on eBay and the likes.

1 person has thanked this post

Want to share your thoughts?

Login here to discuss!